This article was downloaded by: [Pontificia Universidad Catolica de Chile]

On: 22 January 2015, At: 10:23

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Scientific Studies of Reading

Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/hssr20

Modeling the Length Effect: Specifying

the Relation With Visual and
Phonological Correlates of Reading
a a
Madelon van den Boer , Peter F. de Jong & Marleen M. Haentjens-
a
van Meeteren
a
Research Institute of Child Development and Education , University
of Amsterdam
Published online: 05 Oct 2012.

To cite this article: Madelon van den Boer , Peter F. de Jong & Marleen M. Haentjens-van Meeteren
(2013) Modeling the Length Effect: Specifying the Relation With Visual and Phonological Correlates of
Reading, Scientific Studies of Reading, 17:4, 243-256, DOI: 10.1080/10888438.2012.683222

To link to this article: http://dx.doi.org/10.1080/10888438.2012.683222

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
 Scientific Studies of Reading, 17:243–256, 2013
Copyright © 2013 Society for the Scientific Study of Reading
ISSN: 1088-8438 print/1532-799X online
DOI: 10.1080/10888438.2012.683222

Modeling the Length Effect: Specifying the Relation With

Visual and Phonological Correlates of Reading
Madelon van den Boer, Peter F. de Jong,
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

and Marleen M. Haentjens-van Meeteren

Research Institute of Child Development and Education,
University of Amsterdam

Beginning readers’ reading latencies increase as words become longer. This length effect is believed
to be a marker of a serial reading process. We examined the effects of visual and phonological skills
on the length effect. Participants were 184 second-grade children who read 3- to 5-letter words and
nonwords. Results indicated that reading latencies could be decomposed into a length effect and
an overall reading speed. Individual differences in the length effect were predicted by phonological
awareness and visual attention span. Rapid naming accounted only for differences in overall reading
speed.

In the early stages of reading development, the speed of word reading is closely related to the
number of letters in a word. Reading latencies tend to increase as a function of word length. This
length effect is strong for both words and nonwords in young beginning readers and poor readers
but becomes restricted to longer words (i.e., more than six or eight letters) and nonwords in
advanced readers (e.g., De Luca, Barca, Burani, & Zoccolotti, 2008; Hawelka, Gagl, & Wimmer,
2010; Marinus & de Jong, 2010; Spinelli et al., 2005; Ziegler, Perry, Ma-Wyatt, Ladner, &
Schulte-Körne, 2003; Zoccolotti et al., 2005). Generally, a length effect is presumed to occur
when words are processed sequentially, whereas the absence of a length effect (i.e., reading
latencies that are independent of word length) is considered to indicate that words are processed
in parallel. The decrease in the length effect over the course of development is taken to reflect
a gradual shift from serial (e.g., letter-by-letter) word reading, toward the parallel processing of
increasingly larger parts of words. Therefore, the length effect can be seen as a marker of the
reading process (e.g., Ziegler et al., 2003).
The explanation of length effects has been a major focus in several computational models of
reading (Ans, Carbonnel, & Valdois, 1998; Coltheart, Rastle, Perry, Langdon, & Ziegler, 2001;
Perry, Ziegler, & Zorzi, 2007). For example, in the dual route cascaded model (Coltheart et al.,
2001), two reading routes are distinguished: the nonlexical route and the lexical route. In the
lexical route, all letters in a string are activated in parallel, and these letters activate a word’s entry
in the orthographic lexicon. In the nonlexical route, the letters are serially decoded into phonemes

Correspondence should be sent to Madelon van den Boer, Research Institute of Child Development and Education,
University of Amsterdam, P.O. Box 94208,1090 GE Amsterdam, the Netherlands. E-mail: m.vandenboer@uva.nl
 244 VAN DEN BOER, DE JONG, HAENTJENS-VAN MEETEREN

according to grapheme–phoneme conversion rules. Length effects arise when words are not in the
orthographic lexicon and therefore are predominantly read through the nonlexical route. Thus, in
the dual route cascaded model, and other computational models as well (e.g., Ans et al., 1998;
Perry et al., 2007), the length effect reflects the type of reading process that occurs between the
visual presentation of a word and its identification. Although most of these computational models
have not been designed to model the development of reading, individual differences in the length
effect might be regarded as an indication of the development of the reading system (Jackson &
Coltheart, 2001).
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

Currently, there is ample evidence about the cognitive abilities that underlie individual dif-
ferences in reading development, but the majority of this evidence is based on the outcome
of the reading system, the accuracy, or speed of word reading (e.g., de Jong & van der Leij,
1999; Lervåg, Bråten, & Hulme, 2009; Parrila, Kirby, & McQuarrie, 2004; Torgesen, Wagner,
Rashotte, Burgess, & Hecht, 1997). There are very few studies, however, that considered the
relation between these cognitive abilities and individual differences in parameters that reflect the
process of word identification (for important exceptions, see Hawelka & Wimmer, 2005, and
Ziegler et al., 2008). In the current study, we examined the relation between cognitive abilities
and individual differences in the sensitivity to word length.
Phonological-processing skills have been found to be among the major factors underlying indi-
vidual differences in reading ability (e.g., Vellutino, Fletcher, Snowling, & Scanlon, 2004). Three
phonological-processing skills have been distinguished: phonological awareness, rapid naming,
and verbal short-term memory (e.g., Wagner & Torgesen, 1987). Phonological awareness and
rapid naming have repeatedly been reported to predict concurrent (e.g., Vaessen & Blomert, 2010;
Ziegler et al., 2010) as well as future reading skills in all primary school grades (e.g., de Jong &
van der Leij, 1999; Kirby, Parrila, & Pfeiffer, 2003; Landerl & Wimmer, 2008; Lervåg et al., 2009;
McCallum et al., 2006; Parrila et al., 2004; Torgesen et al., 1997). In addition, studies focusing on
reading impairment have shown that reading difficulties tend to be associated with poor naming
speed, poor phonological awareness, or both (e.g., Compton, DeFries, & Olson, 2001; de Jong &
van der Leij, 2003; Lovett, Steinbach, & Frijters, 2000; McBride-Chang & Manis, 1996; Wimmer,
Mayringer, & Landerl, 2002; Wolf & Bowers, 1999; Wolf et al., 2002). The results concerning
verbal short-term memory have been less consistent. Some studies showed that short-term mem-
ory contributed in predicting reading performance, over and above phonological awareness and
rapid naming (e.g., McCallum et al., 2006; Ziegler et al., 2010), whereas in others it was found
that the relation between short-term memory and reading overlapped entirely with the effect of
phonological awareness (e.g., de Jong & van der Leij, 1999; Lervåg et al., 2009; Parrila et al.,
2004).
In short, there is ample evidence that phonological awareness and rapid naming are related
to reading. However, their relation with the length effect is less clear. Phonological awareness
might be related to the length effect through its relation with phonological recoding. According
to the self-teaching hypothesis (e.g., Share, 1995, 1999) beginning readers rely on print-to-sound
translation, or phonological recoding, to read words. Every time a printed word is successfully
translated into a phonological code, the orthographic representation of the word, necessary for
parallel processing, is built and strengthened. However, phonological recoding cannot be suc-
cessful without a basic awareness of phonemes. If phonological awareness is poor, phonological
recoding will be slow and fallible and the connection between a word’s phonological and
orthographic representation will not be established. Accordingly, poor phonological awareness
 MODELING THE LENGTH EFFECT 245

results in poor buildup of orthographic representations, and therefore in continued reliance on a

serial processing strategy, and a remaining sensitivity to word length.
The relation between the rapid naming of letters or digits and the length effect is more specu-
lative. Rapid naming could relate to the activation of letter sounds. To process a word in parallel,
all letters in a word need to be activated within a short timeframe. Although the activation of the
letters itself could be either serial or parallel, it has been suggested by Wolf and Bowers (1999)
among others, that if a reader is slow in identifying letters, all letters in a word would not be
activated fast enough for the integration of individual letters into a whole word representation
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

enabling parallel processing. As a result, the reader would continue to rely on serial processing.
More recently, it has been suggested that in addition to phonological processing skills, the
visual attention span, the number of orthographic units (e.g., letters or syllables) that can be pro-
cessed in parallel, also contributes to reading performance (Valdois et al., 2003; Valdois, Bosse,
& Tainturier, 2004). Theoretically, the visual attention span hypothesis has been grounded in
the multiple-trace memory model (Ans et al., 1998; Bosse, Tainturier, & Valdois, 2007). In this
model, two reading procedures are distinguished that work successively. Through the global pro-
cedure words are read as a whole, based on knowledge of entire words. If identification of a word
cannot be accomplished through the global procedure, the analytic procedure is activated. This
procedure is based on the successive activation of smaller orthographic units, such as syllables or
letters, resulting in length effects. These two reading procedures differ in the type of visual atten-
tion required; the global procedure requires the visual attention span to extend over the whole
letter string, whereas in the analytic procedure visual attention is focused successively on parts
of the input string. Thus, global processing requires a larger visual attention span. If the visual
attention span is too small to cover entire words, these words cannot be processed in parallel, and
serial analysis of the letter string remains the only available reading strategy.
The visual attention span has been shown to contribute to reading performance, independent
from IQ, vocabulary, and phonological processing in typically developing children across pri-
mary school grades (Bosse & Valdois, 2009), as well as in children with reading impairments
(Bosse et al., 2007). Although visual attention span is most often measured as the number of
letters that can be processed in parallel, the same relation with reading has been found for the
parallel processing of nonverbal symbol strings (Lobier, Zoubrinetzky, & Valdois, 2012; Pammer,
Lavis, Hansen, & Cornelissen, 2004, but see Ziegler, Pech-Georgel, Dufau, & Grainger, 2010).
In addition, there is some evidence that visual attention span is related specifically to serial pro-
cessing. Valdois et al. (2006), for example, reported increased activation of visual attentional
brain areas during nonword reading through the analytic procedure. Furthermore, visual attention
span related to the number of eye movements during word reading (Hawelka & Wimmer, 2005),
more specifically to the number of rightward fixations (Prado, Dubois, & Valdois, 2007). The
number of eye movements can be regarded as a parameter of the reading process, similar to the
length effect, as the number of fixations per word indicates whether processing is serial (multiple
fixations within a word) or parallel (a single fixation per word).
From the aforementioned studies it becomes clear that there is ample evidence that both
phonological-processing skills and visual attention span are related to reading ability. However,
to our knowledge, it has not been studied whether phonological and/or visual processing skills
explain variance in the length effect found in reading latencies for words and nonwords. In the
current study we aim to present a model for the length effect that enables us to examine how the
 246 VAN DEN BOER, DE JONG, HAENTJENS-VAN MEETEREN

main phonological-processing skills, phonological awareness, and rapid naming, as well as the
visual attention span, are related to the length effect as a parameter of the reading process.

METHOD

Participants
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

One hundred eighty-four Dutch children (93 boys, 91 girls) from 7 second-grade classes partici-
pated in the study. The children had a mean age of 7 years, 11 months (SD = 5.66 months). At the
time of testing, the children had received approximately 1 year 5 months of reading instruction.
On the One Minute Reading Test (Eén Minuut Test; Brus & Voeten, 1995), a standardized test
of word reading speed with an average of 10 and a standard deviation of 3, the children scored
slightly above average (M = 11.66), with a somewhat smaller standard deviation (SD = 2.53),
indicating that fewer extreme scores were included than in the overall population. All children
had normal or corrected to normal vision.

Measures

Word and nonword reading. The reading task consisted of 45 words and 45 nonwords
varying in length from three to five letters. For each length, 15 monosyllabic words were selected
from a corpus of child literature (Schrooten & Vermeer, 1994). Across lengths the words were
matched on onset (i.e., the first phoneme) and frequency. The words ranged in frequency to reflect
the variation in the words children encounter (Mdn = 23, range = 1–148). In comparison, for all
monosyllabic words in the database the median frequency was 22, with a range of 1 to 591
(excluding 10% extremely high-frequent words).
Nonwords were created by interchanging onsets and rhymes of the words. For example,
the words drift, front, and kramp (meaning urge, front, and cramp, respectively) were used
to create the nonwords “dront,” “framp,” and “krift.” Therefore, words and nonwords of each
length were matched on onset and consonant-vowel structure. When the created nonword was
unpronounceable or also a Dutch word, one letter was changed in the rhyme.
The reading task (as well as the verbal short-term memory task and the visual attention
span task mentioned next) was programmed in E-prime Version 1.0 (Schneider, Eschman, &
Zuccolotto, 2002). Words and nonwords were presented one by one in the middle of a laptop
screen (14.1 in.; 35.8 cm) in 72-point Arial font. A plus sign presented for 750 ms focused atten-
tion. Then the word or nonword appeared, and children were asked to read it aloud as quickly
and accurately as possible. A voice key registered naming latencies from the onset of stimulus
presentation until the onset of the response. The experimenter registered for each trial whether
the response was correct and the latency was valid, by pressing the corresponding button on
a response box (correct and valid, incorrect, or invalid). Words and nonwords were presented
in blocks separated by a fixed break of 1.5 min. The order of word and nonword reading was
counterbalanced across the children.

Vocabulary. The Vocabulary subtest from the RAKIT intelligence test battery (Bleichrodt,
Drenth, Zaal, & Resing, 1984) was used to assess children’s vocabulary. The experimenter read
 MODELING THE LENGTH EFFECT 247

aloud a word, and children were instructed to choose from among four alternatives the picture
that best matched the word. A total of 45 items were administered. The score consisted of the
number of items correct.
Nonverbal IQ. Raven’s Standard Progressive Matrices test (Raven, 1960) was used to assess
nonverbal IQ. Children were asked to identify out of six or eight alternatives the element that fitted
the missing position in a pattern. Children were instructed to work individually for 45 min and
complete as many items as possible. The score consisted of the number of items correct.
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

Verbal short-term memory. To assess verbal short-term memory, a letter memory span task
was used (e.g., Johnston, Rugg, & Scott, 1987). Children were presented with letter sequences
increasing in length from two to seven letters. Letters were presented visually, one by one on
a laptop screen. Each letter was presented for 750 ms in bold 24-point Arial font. Letters were
separated by a blank screen presented for 500 ms. When the entire sequence had been presented,
children were asked to repeat the letters aloud in the order of presentation. Children were pre-
sented with three sequences of each length. The task was discontinued after two consecutive
errors within the same sequence length. The score consisted of the number of sequences repeated
correctly.
Phonological awareness. An elision task was used to assess phonological awareness (de
Jong & van der Leij, 2003). The experimenter read aloud a nonword, and children were asked
to repeat the nonword completely. Next, the experimenter repeated the nonword and named
a phoneme to be deleted. Children were asked to repeat the nonword without this phoneme.
Children completed a total of 27 items. The first 18 items required a single phoneme to be deleted
(e.g., “tral” without “r”), whereas in the last nine items the phoneme to be deleted was included
twice (e.g., “gepgral” without “g”). The score consisted of the number of correct responses.
Rapid naming. To assess rapid naming, children were presented with a sheet with five lines
of 10 digits each. The digits were a sequence of the digits 1, 3, 5, 6, and 8, each repeated 10 times
in random order. Children were asked to name the digits aloud as quickly as possible. The time
needed to name all digits on the sheet was recorded. The score consisted of the naming latency
in milliseconds per digit.
Visual attention span. Following Valdois et al. (2003) the whole report task was used to
assess children’s visual attention span. Stimuli consisted of 20 five-letter strings (e.g., R H S D
M). The strings were created from the consonants B, D, F, H, L, M, P, R, S, and T. Each consonant
was repeated in 10 strings, twice in each of the five letter positions. Letters were presented in
bold 24-point Arial font. The strings were presented for 200 ms. Children were asked to repeat as
many letters of the string as possible. Trials were preceded by a plus sign presented for 1,000 ms
to focus attention. The score consisted of the number of letters repeated correctly (from a total of
100), in its proper position within the string. For example, if a child would answer “R H S M”
in response to “R H S D M,” this child would score 3 on this particular item, because M was the
fifth, not the fourth letter in the string.1

1 In this score both identity and location of the reported letters were taken into account. This score differs from

the score used in previous studies (e.g., Valdois et al., 2003), which was based on the total number of letters that was
identified correctly. In our score it was also taken into account whether order of the letters was correct, because in reading
 248 VAN DEN BOER, DE JONG, HAENTJENS-VAN MEETEREN

Procedure

The vocabulary and nonverbal IQ tasks were administered during a classroom session of about
45 min each. The other tasks were administered in a fixed order to each child individually in two
sessions of approximately 30 min.

Analysis
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

Structural equation modeling was used to model the relation between the length effect and the
cognitive correlates of reading. The model was specified in two steps. First, the relation between
the number of letters in a word or nonword and its reading latency was specified as a latent growth
model. Such a model consists of an intercept, or initial status, and a slope, or growth rate over
time (Kline, 2011). In our model, however, we estimated the slope not over time, but over word
length. The intercept was specified as the reading speed at the first measurement point, namely,
three-letter words or nonwords, and the slope as the additional time needed to read words and
nonwords that contain one or two more letter(s). Because an intercept and slope were estimated
simultaneously but separately for words and nonwords, the resulting model can be considered a
parallel latent “growth” model, with two intercepts and two slopes (Kaplan, 2009). In the second
step, the correlates of reading were added as concurrent predictors of the length effect. Direct
effects were specified from each correlate to the intercepts and slopes in the length effect model.
Analyses were conducted within Mplus version 5.21 (Muthén & Muthén, 2009). Parameter
estimates were obtained by full information maximum likelihood estimation. To evaluate model
fit we used the chi-square statistic of overall goodness of fit, the root mean square error of
approximation (RMSEA), the comparative fit index (CFI), and the standardized root mean square
residual (SRMR; Kline, 2011). A significant chi-square indicates that the model does not fit the
data, whereas p values considerably larger than .05 indicate exact fit (Hayduk, 1996). For the
RMSEA, values below .05 indicate close fit, values below .08 indicate satisfactory fit, and values
over .10 indicate poor fit (Browne & Cudeck, 1993). A CFI larger than .95 in combination with
an SRMR below .08 indicates good fit (Hu & Bentler, 1999). To test the difference in model
fit between two hierarchical models, a chi-square test of the difference in chi-squares and the
difference in degrees of freedom was used (Kline, 2011).

RESULTS

Data Cleaning

For the word and nonword latencies, analyses were conducted on correct and valid trials only.
Trials were excluded from analysis if the response was incorrect (8.8%) or if the voice key was
not validly triggered (6.2%). In addition, latencies of less than 350 ms or more than 6,000 ms
(1.3%), as well as latencies 3 standard deviations below or above a participant’s mean (0.9%)
were considered invalid and removed. For each child valid and correct trials were averaged for

it is important to perceive letters in the correct order in which they appear. However, the correlation between the score
with and without taking order into account was .88, and the use of identity only scores in our analyses gave similar results.
 MODELING THE LENGTH EFFECT 249

each of the six Lexicality × Length conditions. Mean scores were coded as missing when less
than 60% of the trials for that condition were valid. Three children were excluded from further
analysis due to missing scores in more than two of the six conditions.
Scores on the nonverbal IQ task were recalculated. Due to a mistake of one of the experi-
menters 67 children completed only the first 36 items of the test. However, for the children that
completed as many items as possible, the score over the first 36 items correlated highly with the
total score (r = .925, p < .01). Therefore, for all children the nonverbal IQ score was calculated
over the first 36 items, and this score was used in the analyses reported in the following sections.
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

Scores on the correlates of reading were all normally distributed. One univariate outlier (i.e.,
a score of more than 3 standard deviations above or below the mean) was identified for the
verbal short-term memory task, and one for the vocabulary task. Furthermore, based on their
score pattern, six children were identified as a multivariate outlier, using Mahalanobis distance
with a significance level of .001 (Tabachnick & Fidell, 2001). These children were excluded from
further analysis. Taken together, the reported analyses were based on data from 175 children
(90 boys, 85 girls) with a mean age of 7 years, 11 months (SD = 5.62 months).

Length Effects

The mean latencies and error rates for the words and nonwords of each length are reported in
Table 1. The error rates were below 10% in all conditions and were therefore not analyzed. The
distributions of the latencies were positively skewed. To normalize the data, a log transformation
was used. This transformation also corrects for proportional increases in naming latencies for
words and nonwords of increasing length (Keene, 1995). Differences in the length effect are
affected by the mean latency, because initial differences in latencies for three-letter words and
nonwords become larger at four- and five-letter words and nonwords. In other words, a larger
increase in latencies from three- to four-letter words and nonwords when latencies to three-letter
words and nonwords are long might be proportionally equal to a smaller increase with shorter
mean latencies. Our main interest, however, is in length effects that are disproportionate to the
mean latencies. Log-transformed scores control for proportional increases and provide a purer
estimate of the length effect. The means after this transformation are also presented in Table 1.

TABLE 1
Mean Reaction Times (RT) in Milliseconds, Mean RTs After Log Transformation, and Mean Error
Rates in Percentages (and Standard Deviations)

RT (in ms) Error Rates RT (Log Transformed)

Words
3 letters 828 (377) 4.7 (7.6) 6.65 (.34)
4 letters 982 (571) 8.5 (10.2) 6.78 (.44)
5 letters 1, 106 (726) 7.7 (9.7) 6.87 (.48)
Nonwords
3 letters 963 (443) 5.9 (8.2) 6.79 (.38)
4 letters 1, 152 (658) 10.0 (10.9) 6.93 (.47)
5 letters 1, 242 (744) 9.9 (10.7) 6.99 (.50)
 250 VAN DEN BOER, DE JONG, HAENTJENS-VAN MEETEREN

The log-transformed scores were subjected to an analysis of variance (ANOVA). The main
effects of lexicality, F(1, 155) = 125.89, p < .001, ηp 2 = 0.45, and length, F(2, 154) = 88.40,
p < .001, ηp 2 = 0.53, were significant. For both words and nonwords the length effect was
significant from three to four letters, F(1, 155) = 82.24, p < .001, ηp 2 = 0.35, and F(1, 155) =
103.02, p < .001, ηp 2 = 0.40, respectively, and from four to five letters, F(1, 155) = 87.90, p
< .001, ηp 2 = 0.36, and F(1, 155) = 37.21, p < .001, ηp 2 = 0.19, respectively. The interaction
between lexicality and length did not reach significance.
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

Modeling the Length Effect

To describe the relation between word and nonword length and reading latency, a parallel latent
growth model was fitted to the variance–covariance matrix of three-, four-, and five-letter word
and nonword latencies. The length effect was modeled as a linear process, with increases in
latencies from three- to four-, and from four- to five-letter words and nonwords constrained to
be equal. This model provided a poor fit to the data, χ 2 (7) = 27.10, p = .000, RMSEA = .128
[.079–.181], CFI = .990, SRMR = .128. To improve model fit, nonlinear length effects were
considered by allowing the loadings of the four-letter words and nonwords on the slope factors to
be freely estimated. This model provided a good fit to the data, χ 2 (5) = 2.46, p = .782, RMSEA
= .000 [.000–.069], CFI = 1.000, SRMR = .045, and was also a significant improvement over
the previous model, χ 2 (2) = 24.64, p < .001.
For all four factors, the variances (intercept words = .111, slope words = .011, intercept
nonwords = .140, slope nonwords = .010) differed significantly from zero (p < .001), indicat-
ing that there was individual variation in both intercepts and slopes. Correlations between the
intercept and slope for words (r = .608) and nonwords (r = .477) were moderate. The intercepts
for words and nonwords were strongly correlated (r = .904), as were the slopes (r = .757). The
correlations between the intercept for words and the slope for nonwords (r = .704) and between
the intercept for nonwords and the slope for words (r = .503) were also moderate to high.

Cognitive Correlates of the Length Effect

The means, standard deviations, and range of the scores on the cognitive correlates are presented
in Table 2, as well as the correlations with word and nonword reading. Rapid naming latencies
were log transformed in the same way as the word and nonword latencies. The relations between
the cognitive correlates of reading and the intercepts and slopes were examined. Direct effects
were specified for each predictor on each of the four factors in the length effect model. The
model provided a good fit to the data, χ 2 (17) = 15.43, p = .565, RMSEA = .000 [.000–.062],
CFI = 1.000, SRMR = .025. A graphical representation of the model, including standardized
parameter estimates of the significant direct effects, is presented in Figure 1. The standardized
direct effects of the control variables vocabulary, nonverbal IQ, and verbal short-term memory
were not significant (range = −.11 to .05).
Individual differences in the intercepts, representing the reading speed for three-letter words
and nonwords, were related to rapid naming, phonological awareness, and visual attention span.
Individual differences in the slopes for words and nonwords, representing the length effect, were
associated with phonological awareness and visual attention span but not with rapid naming.
Nonverbal intelligence, vocabulary, and verbal short-term memory did not explain any additional
 MODELING THE LENGTH EFFECT 251

TABLE 2
Means, Standard Deviations, and Correlations for Reading Variables and Cognitive Correlates of Reading

Variable 1 2 3 4 5 6 7 8

Reading speed
1 Words 1
2 Nonwords .937∗∗ 1
Cognitive correlates
3 Vocabulary −.020 .001 1
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

4 Nonverbal IQ −.130 −.087 .184∗ 1

5 Short-term −.272∗∗ −.233∗∗ .044 .226∗∗ 1
memory
6 Phonological −.418∗∗ −.409∗∗ −.008 .252∗∗ .359∗∗ 1
awareness
7 Rapid naming .364∗∗ .380∗∗ .125 .126 −.115 −.303∗∗ 1
8 Visual attention −.353∗∗ −.354∗∗ .010 .168∗ .335∗∗ .306∗∗ −.203∗∗ 1
span

M 6.78 6.91 27.35 24.11 6.89 15.75 6.42 53.90

SD .42 .44 3.81 5.47 1.65 5.40 .22 12.11
Range 6.15–8.19 6.16–8.12 18–36 7–34 3–11 1–27 5.94–7.07 31–95

Note. The full correlation matrix (i.e., the input for the structural equation models) can be obtained from the
corresponding author.
∗ p < .05. ∗∗ p < .01.

FIGURE 1 Model of the length effect for words and nonwords. This
figure contains only the significant effects (α = .05) on the intercepts and
slopes.

variance in any of the factors. Taken together, the visual and phonological abilities explained
27.6% of the variance in the intercept for words, 28.9% in the intercept for nonwords, 20.9% in
the slope for words, and 15.9% in the slope for nonwords.
Of interest, the direct effects of these cognitive correlates on the length effect factors looked
quite similar. Therefore, we examined whether the effects on words and nonwords, and the effects
 252 VAN DEN BOER, DE JONG, HAENTJENS-VAN MEETEREN

on intercepts and slopes, respectively, were equal. First, the effects of the cognitive correlates
on the intercepts for words and nonwords were constrained to be equal, as were the effects on
the slopes. This model provided a good fit to the data, χ 2 (29) = 27.13, p = .565, RMSEA =
.000 [.000–.053], CFI = 1.000, SRMR = .082, and did not differ significantly from the previous
model, χ 2 (12) = 11.70, p = .470. Second, the effects of the cognitive correlates on the intercept
and slope for words were constrained to be equal, as were the effects on the intercept and slope for
nonwords. This model provided a poor fit to the data, χ 2 (29) = 71.88, p < .001, RMSEA = .092
[.065–.119], CFI = .979, SRMR = .225, and was significantly worse than the previous model,
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

χ 2 (12) = 56.45, p < .001. Even with rapid naming, which had an effect on the intercepts only,
excluded from the equality constraints, the model remained significantly worse than the previous
model, χ 2 (10) = 23.14, p < .05. In sum, the direct effects of the predictors of reading differed
between intercepts and slopes but did not differ between words and nonwords.

DISCUSSION

In the current study we examined how visual and phonological correlates of reading ability are
related to the length effect, a marker of the reading process. As expected, we found that second-
grade beginning readers showed a length effect for both words and nonwords. This result, in
line with previous studies (e.g., Marinus & de Jong, 2010; Zoccolotti et al., 2005), indicates that
these readers do not process letter strings in parallel but predominantly rely on a serial processing
strategy. Nonwords were read more slowly than words, but the interaction between length and
lexicality did not reach significance. Although larger length effects for nonwords than for words
are predicted by several models of reading (e.g., Ans et al., 1998; Coltheart et al., 2001), the
absence of this interaction in children has been reported previously (e.g., Ziegler et al., 2003).
The detection of the interaction could depend on the frequency of the words. High-frequency
words will most likely be in the orthographic lexicon of children, enabling parallel processing.
For words of lower frequency, orthographic representations may not have been built yet, resulting
in a length effect.
The main aim of this study was to model the length effect and its relations with phonological
and visual correlates of reading. Our results indicated that reading latencies of words and
nonwords could be decomposed into an intercept and a slope. If the intercepts and slopes would
reflect similar aspects of the reading process, the correlation between the word and nonword inter-
cept and slope would be close to one. However, we found that these factors were only moderately
related, which indicates that the intercepts and slopes reflect partly different processes.
The slopes represent the length effect, or the degree of serial processing. If processing is serial,
there is an increase in reading latencies with each additional letter, whereas the slope will be close
to zero if words are mainly processed in parallel. Interpretation of the intercepts, however, is not
as straightforward. Because the intercepts were estimated based on reading latencies to three-
letter words and nonwords, which consist of multiple letters, the intercepts may partly reflect
a serial reading process. If readers rely on a serial processing strategy, this type of processing
can also be expected for three-letter words and nonwords. However, the intercepts also represent
overall reading speed of words and nonwords, irrespective of length. These findings are consistent
with the developmental trends described by Spinelli et al. (2005) and Zoccolotti et al. (2005),
who showed that over the course of development, children’s word reading is characterized by a
 MODELING THE LENGTH EFFECT 253

decrease in the length effect, but also by an increase in overall reading speed, regardless of word
length. Our model indicates that although, in general, slower reading times are associated with
a greater reliance on serial processing, these aspects are distinguishable, and represent different
aspects of the reading process.
The relations of the intercepts and slopes with the correlates of reading further support the
dissociation between overall reading speed and the degree of serial processing. The overall read-
ing speed of words and nonwords, the intercept, was predicted by phonological awareness, rapid
naming, and visual attention span. This is in line with previous studies (e.g., Bosse & Valdois,
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

2009; Vaessen & Blomert, 2010; Ziegler et al., 2010), that showed concurrent relations of these
cognitive skills with reading ability. The length effect, the slope, however, was also predicted
by phonological awareness and visual attention span, but not by rapid naming. The relation
of phonological awareness and visual attention span with the slopes is in line with the results
reported by Hawelka and Wimmer (2005), who found that these skills relate to the degree to
which words and nonwords are processed serially.
As in previous studies (e.g., Powell, Stainthorp, Stuart, Garwood, & Quinlan, 2007), the effect
of rapid naming on reading was largely independent from the effect of phonological awareness.
A novel finding was, however, that rapid naming and phonological awareness affected different
aspects of reading. Rapid naming accounted for individual differences in the intercepts only,
which indicates that rapid naming, in contrast to phonological awareness, is specifically related
to overall reading speed. The effect of rapid naming was the same for words and nonwords.
A similar finding was reported by Moll, Fussenegger, Willburger, and Landerl (2009), suggesting
that rapid naming cannot be particularly involved in the acquisition of orthographic knowledge.
The current results suggest that rapid naming is related to an aspect of reading that is shared in
the processing of all letter strings, irrespective of length or lexicality.
The effects of rapid naming are especially interesting, because our results suggest that children
with poor phonological awareness and/or visual attention span can be expected to be slow readers
who continue to rely on serial processing, whereas children with poor rapid naming skills are most
likely slow readers who do shift toward a parallel-processing strategy and who thus show a dif-
ferent kind of reading impairment. This finding could be interesting in light of the double-deficit
hypothesis (e.g., Wolf & Bowers, 1999). According to this hypothesis, phonological awareness
and rapid naming represent two separate deficits underlying reading impairment. Our results
also indicate this unique contribution of phonological awareness and rapid naming to reading
(dis)ability, and they suggest that these skills might be associated with different types of reading
problems.
As predicted, visual attention span had an effect on the degree of serial processing for both
words and nonwords. It has been argued by Bosse and Valdois (2009) that visual attention span
should not be perceived as a measure of verbal short-term memory. To examine this argument, we
included the letter memory span task. Both the form of presentation and the stimuli of this task
were very similar to the visual attention span. However, we found that verbal short-term memory
did not have an independent effect on the intercepts or on the slopes. This finding is in accordance
with studies that indicated that the relation between short-term memory and reading is shared
with the other phonological-processing variables (e.g., de Jong & van der Leij, 1999; Parrila et al.,
2004). Visual attention span, however, even with verbal short-term memory included in the model,
had a significant effect on all aspects of the reading process. These results support the claim
made by Bosse and Valdois (2009) that the visual attention span does not reflect general or verbal
 254 VAN DEN BOER, DE JONG, HAENTJENS-VAN MEETEREN

short-term memory, but rather multielement parallel visual processing. Although it is still debated
whether visual attention span concerns a purely visual skill (Lobier et al., 2012), or visual to
phonological mapping (Ziegler et al., 2010), our results clearly indicate that visual attention span
predicts reading latencies independent from phonological-processing skills. If visual attention
span reflects visual to phonological mapping, a stronger relation might have been expected with
rapid naming, a skill that is also associated with orthography to phonology mappings. Therefore,
our results are more in favor of a purely visual interpretation of visual attention span.
The amount of variance in overall reading speed and length effects explained by the model
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

ranged from 15.9 to 28.9%, which is in line with previous studies that included similar predictors
of reading (e.g., Bosse & Valdois, 2009; de Jong & van der Leij, 1999). Although a substantial
amount of variance is explained, a large part remains unexplained. It should be noted, however,
that a discrete naming task, as used in the current study, differs from most standardized tests, in
which words are presented in a list. For example, recent findings of de Jong (2011) indicate that
discrete rather than serial rapid naming could explain additional variance in a discrete naming
task.
We explicitly modeled individual differences in the length effect to investigate the relation
between cognitive correlates of reading and this specific marker of the developing reading pro-
cess. Phonological awareness and visual attention span were both related to the length effect and
therefore to a serial processing strategy. Rapid naming, in contrast, was related not to the length
effect but to aspects of reading that are independent of word length. Future studies could fur-
ther specify the relations between reading and cognitive correlates by determining the specific
relations of visual and phonological skills with parameters of the reading system.

REFERENCES

Ans, B., Carbonnel, S., & Valdois, S. (1998). A connectionist multiple-trace memory model for polysyllabic word reading.
Psychological Review, 105, 678–723.
Bleichrodt, N., Drenth, P. J. D., Zaal, J. N., & Resing, W. C. M. (1984). Revisie Amsterdamse Kinder Intelligentie Test
[Revised Amsterdam Child Intelligence Test]. Lisse, the Netherlands: Swets & Zeitlinger.
Bosse, M. L., Tainturier, M. J., & Valdois, S. (2007). Developmental dyslexia: The visual attention span deficit hypothesis.
Cognition, 104, 198–230.
Bosse, M. L., & Valdois, S. (2009). Influence of the visual attention span on child reading performance: A cross-sectional
study. Journal of Research in Reading, 32, 230–253.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.),
Testing structural equation models (pp. 136–162). Newbury Park, CA: Sage.
Brus, B., & Voeten, B. (1995). Eén minuut test vorm A en B. Verantwoording en handleiding [One-minute-test manual].
Lisse, the Netherlands: Swets & Zeitlinger.
Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). DRC: A dual route cascaded model of visual word
recognition and reading aloud. Psychological Review, 108, 204–256.
Compton, D. L., DeFries, J. C., & Olson, R. K. (2001). Are RAN- and phonological awareness-deficits additive in children
with reading disabilities? Dyslexia, 7, 125–149.
de Jong, P. F. (2011). What discrete and serial rapid automatized naming can reveal about reading. Scientific Studies of
Reading, 15, 314–337.
de Jong, P. F., & van der Leij, A. (1999). Specific contributions of phonological abilities to early reading acquisition:
Results from a Dutch latent variable longitudinal study. Journal of Educational Psychology, 91, 450–476.
de Jong, P. F., & van der Leij, A. (2003). Developmental changes in the manifestation of a phonological deficit in dyslexic
children learning to read a regular orthography. Journal of Educational Psychology, 95, 22–40.
 MODELING THE LENGTH EFFECT 255

De Luca, M., Barca, L., Burani, C., & Zoccolotti, P. (2008). The effect of word length and other sublexical, lexical, and
semantic variables on developmental reading deficits. Cognitive and Behavioral Neurology, 21, 227–235.
Hawelka, S., Gagl, B., & Wimmer, H. (2010). A dual-route perspective on eye movements of dyslexic readers. Cognition,
115, 367–379.
Hawelka, S., & Wimmer, H. (2005). Impaired visual processing of multi-element arrays is associated with increased
number of eye movements in dyslexic reading. Vision Research, 45, 855–863.
Hayduk, L. A. (1996). LISREL issues, debates and strategies. Baltimore, MD: Johns Hopkins University Press.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus
new alternatives. Structural Equation Modeling, 6, 1–55.
Jackson, N. E., & Coltheart, M. (2001). Routes to reading success and failure: Toward an integrated cognitive psychology
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

of atypical reading. New York, NY: Taylor & Francis.

Johnston, R. S., Rugg, M. D., & Scott, T. (1987). Phonological similarity effects, memory span and developmental reading
disorders: The nature of the relationship. British Journal of Psychology, 78, 205–211.
Kaplan, D. (2009). Structural equation modeling: Foundations and extensions (2nd ed.). Thousand Oaks, CA: Sage.
Keene, O. N. (1995). The log transformation is special. Statistics in Medicine, 14, 811–819.
Kirby, J. R., Parrila, R., & Pfeiffer, S. (2003). Naming speed and phonological awareness as predictors of reading
development. Journal of Educational Psychology, 95, 453–464.
Kline, R. B. (2011). Principles and practice of structural equation modeling (3rd ed.). New York, NY: Guilford.
Landerl, K., & Wimmer, H. (2008). Development of word reading fluency and spelling in a consistent orthography: An
8-year follow-up. Journal of Educational Psychology, 100, 150–161.
Lervåg, A., Bråten, I., & Hulme, C. (2009). The cognitive and linguistic foundations of early reading development: A
Norwegian latent variable longitudinal study. Developmental Psychology, 45, 764–781.
Lobier, M., Zoubrinetzky, R., & Valdois, S. (2012). The visual attention span deficit in dyslexia is visual and not verbal.
Cortex, 48, 768–773.
Lovett, M. W., Steinbach, K. A., & Frijters, J. C. (2000). Remediating the core deficits of developmental reading disability:
A double-deficit perspective. Journal of Learning Disabilities, 33, 334–358.
Marinus, E., & de Jong, P. F. (2010). Variability in the word-reading performance of dyslexic readers: Effects of letter
length, phoneme length and digraph presence. Cortex, 46, 1259–1271.
McBride-Chang, C., & Manis, F. R. (1996). Structural invariance in the associations of naming speed, phonological
awareness, and verbal reasoning in good and poor readers: A test of the double deficit hypothesis. Reading and
Writing: An Interdisciplinary Journal, 8, 323–339.
McCallum, R. S., Bell, S. M., Wood, M. S., Below, J. L., Choate, S. M., & McCane, S. J. (2006). What is the role of
working memory in reading relative to the big three processing variables (orthography, phonology, and rapid naming)?
Journal of Psychoeducational Assessment, 24, 243–259.
Moll, K., Fussenegger, B., Willburger, E. & Landerl, K. (2009). RAN is not a measure of orthographic processing.
Evidence from the asymmetric German orthography. Scientific Studies of Reading, 13, 1–25.
Muthén, L. K., & Muthén, B. O. (2009). Mplus (Version 5.21) [Computer software]. Los Angeles, CA: Muthén & Muthén.
Pammer, K., Lavis, R., Hansen, P., & Cornelissen, P. L. (2004). Symbol-string sensitivity and children’s reading. Brain
and Language, 89, 601–610.
Parrila, R., Kirby, J. R., & McQuarrie, L. (2004). Articulation rate, naming speed, verbal short-term memory, and
phonological awareness: Longitudinal predictors of early reading development? Scientific Studies of Reading, 8, 3–26.
Perry, C., Ziegler, J. C., & Zorzi, M. (2007). Nested incremental modeling in the development of computational theories:
The CDP model of reading aloud. Psychological Review, 114, 273–315.
Powell, D., Stainthorp, R., Stuart, M., Garwood, H., & Quinlan, P. (2007). An experimental comparison between rival
theories of rapid automatized naming performance and its relationship to reading. Journal of Experimental Child
Psychology, 98, 46–68.
Prado, C., Dubois, M., & Valdois, S. (2007). The eye movements of dyslexic children during reading and visual search:
Impact of the visual attention span. Vision Research, 47, 2521–2530.
Raven, J. C. (1960). Standard progressive Matrices: Sets A, B, C, D and E. London, UK: Lewis & Co.
Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-prime: User’s guide. Pittsburgh, PA: Psychology Software Inc.
Schrooten, W., & Vermeer, A. (1994). Woorden in het basisonderwijs: 15.000 woorden aangeboden aan leerlingen
[Words in primary education: 15.000 words presented to students]. Tilburg, the Netherlands: Tilburg University
Press.
 256 VAN DEN BOER, DE JONG, HAENTJENS-VAN MEETEREN

Share, D. L. (1995). Phonological recoding and self-teaching: Sine qua non of reading acquisition. Cognition, 55,
151–218.
Share, D. L. (1999). Phonological recoding and orthographic learning: A direct test of the self-teaching hypothesis.
Journal of Experimental Child Psychology, 72, 95–129.
Spinelli, D., de Luca, M., Filippo, G. D., Mancini, M., Martelli, M., & Zoccolotti, P. (2005). Length effect in word naming
in reading: Role of reading experience and reading deficit in Italian readers. Developmental Neuropsychology, 27,
217–235.
Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Boston, MA: Allyn & Bacon.
Torgesen, J. K., Wagner, R. K., Rashotte, C. A., Burgess, S., & Hecht, S. (1997). Contributions of phonological awareness
and rapid automatic naming ability to the growth of word-reading skills in second- to fifth-grade children. Scientific
Downloaded by [Pontificia Universidad Catolica de Chile] at 10:23 22 January 2015

Studies of Reading, 1, 161–185.

Vaessen, A., & Blomert, L. (2010). Long-term cognitive dynamics of fluent reading development. Journal of Experimental
Child Psychology, 105, 213–231.
Valdois, S., Bosse, M. L., Ans, B., Carbonnel, S., Zorman, M., David, D., & Pellat, J. (2003). Phonological and visual
processing deficits can dissociate in developmental dyslexia: Evidence from two case studies. Reading and Writing:
An Interdisciplinary Journal, 16, 541–572.
Valdois, S., Bosse, M. L., & Tainturier, M. J. (2004). The cognitive deficits responsible for developmental dyslexia:
Review of evidence for a selective visual attentional disorder. Dyslexia, 10, 339–363.
Valdois, S., Carbonnel, S., Juphard, A., Baciu, M., Ans, B., Peyrin, C., & Segebarth, C. (2006). Polysyllabic pseudo-word
processing in reading and lexical decision: Converging evidence from behavioral data, connectionist simulations and
functional MRI. Brain Research, 1085, 149–162.
Vellutino, F. R., Fletcher, J. M., Snowling, M. J., & Scanlon, D. M. (2004). Specific reading disability (dyslexia): What
have we learned in the past four decades? Journal of Child Psychology and Psychiatry, 45, 2–40.
Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological processing and its causal role in the acquisition of
reading skills. Psychological Bulletin, 101, 192–212.
Wimmer, H., Mayringer, H., & Landerl, K. (2000). The double-deficit hypothesis and difficulties in learning to read a
regular orthography. Journal of Educational Psychology, 92, 668–680.
Wolf, M., & Bowers, P. G. (1999). The double-deficit hypothesis for the developmental dyslexias. Journal of Educational
Psychology, 91, 415–438.
Wolf, M., O’Rourke, A.G., Gidney, C., Lovett, M., Cirino, P., & Morris, R. (2002). The second deficit: An investigation
of the independence of phonological and naming-speed deficits in developmental dyslexia. Reading and Writing: An
Interdisciplinary Journal, 15, 43–72.
Ziegler, J. C., Bertrand, D., Toth, D., Csépe, V., Reis, A., Faísca, L., . . . Blomert, L. (2010). Orthographic depth and its
impact on universal predictors of reading: A cross-language investigation. Psychological Science, 21, 554–559.
Ziegler, J. C., Castel, C., Pech-Georgel, C., George, F., Alaria, F. X., & Perry, C. (2008). Developmental dyslexia and the
dual route model of reading: Simulating individual differences and subtypes. Cognition, 107, 151–178.
Ziegler, J. C., Pech-Georgel, C., Dufau, S., & Grainger, J. (2010). Rapid processing of letters, digits, and symbols: What
purely visual-attentional deficit in developmental dyslexia? Developmental Science, 13, F8–F14.
Ziegler, J. C., Perry, C., Ma-Wyatt, A., Ladner, D., & Schulte-Körne, G. (2003). Developmental dyslexia in different
languages: Language specific or universal? Journal of Experimental Child Psychology, 86, 169–193.
Zoccolotti, P., de Luca, M., di Pace, E., Gasperini, F., Judica, A., & Spinelli, D. (2005). Word length effect in early reading
and in developmental dyslexia. Brain and Language, 93, 369–373.